Ultrasound system with deep learning network for image artifact identification and removal

Abstract

An ultrasound system with a deep learning neural network feature is used to eliminate haze artifacts in B mode images of the carotid artery by analysis of orthogonal information. In a described implementation the orthogonal information comprises the structural information of a B mode image and motion information of the same field of view as that of the B mode image. In another embodiment the neural network reduces haze artifacts by reducing TGC gain at the depth of artifacts.

What is claimed is: 
     
       1. An ultrasonic diagnostic imaging system for improving the image quality of ultrasound images using deep learning comprising:
 an ultrasound probe adapted to acquire ultrasound image signals; 
 a B mode image processor, coupled to the probe, which is adapted to produce B mode ultrasound images; 
 a neural network model, coupled to receive the B mode ultrasound images, and adapted to identify artifacts in a blood vessel in the B mode ultrasound images; 
 time gain control circuitry, coupled to receive the ultrasound image signals, wherein the time gain control circuitry is adapted to be responsive to the identification of artifacts to reduce gain at an image depth of identified artifacts; and 
 a display adapted to display the B mode images with reduced artifact content. 
 
     
     
       2. The ultrasonic diagnostic imaging system of  claim 1 , further comprising a motion detector, coupled to the probe, which is adapted to produce information about motion in the region imaged by the B mode ultrasound images, wherein the neural network model is further adapted to use information about motion in the identification of artifacts. 
     
     
       3. The ultrasonic diagnostic imaging system of  claim 2 , further comprising an artifact filter, coupled to receive the B mode images and adapted to be responsive to the neural network model to reduce artifacts identified in the B mode ultrasound images. 
     
     
       4. The ultrasonic diagnostic imaging system of  claim 3 , further comprising a correlator, responsive to analysis of B mode images by the neural network model and to the information about motion, and having an output coupled to the artifact filter adapted to control the artifact filter. 
     
     
       5. The ultrasonic diagnostic imaging system of  claim 3 , wherein the neural network model is further adapted to produce a confidence factor for display to a user. 
     
     
       6. The ultrasonic diagnostic imaging system of  claim 1 , wherein the neural network model is further adapted to recognize anatomy in the B mode image. 
     
     
       7. The ultrasonic diagnostic imaging system of  claim 2 , wherein the motion detector further comprises a Doppler processor. 
     
     
       8. The ultrasonic diagnostic imaging system of  claim 7 , wherein the Doppler processor is further configured to operate with ensemble lengths shorter that six samples. 
     
     
       9. The ultrasonic diagnostic imaging system of  claim 7 , wherein the Doppler processor is further configured to operate with ensembles acquired by multiline reception. 
     
     
       10. The ultrasonic diagnostic imaging system of  claim 2 , wherein the motion detector is further configured to operate by speckle tracking. 
     
     
       11. The ultrasonic diagnostic imaging system of  claim 1 , wherein the neural network model is further coupled to the time gain control circuitry, and adapted to communicate to the time gain control circuitry the image depth of an identified artifact. 
     
     
       12. The ultrasonic diagnostic imaging system of  claim 1 , wherein the neural network model is further adapted to reduce TGC gain and analyze a reacquired ultrasound image for artifacts in an iterative manner.